Cognitive training based on EEG-neurofeedback to improve working memory : a research study on healthy volunteers with an outlook on preclinical Alzheimer's disease

Abstract

Working memory (WM) has been associated with alpha suppression which can be enhanced by reward expectation. The role of WM is relevant when dealing with neurodegenerative disorders, such as Alzheimer’s disease (AD) causing progressive impairment of WM and underlying oscillations. Thus, understanding these processes would empower AD interventions. This thesis investigated whether providing individuals real-time neurofeedback (NF) about their ongoing oscillations can improve WM under high or low reward expectancies. In two double-blind experiments, sixty participants were trained over 5-days to suppress alpha power while receiving real-time or a control NF during a monetary rewarded delayed match-to-sample task (DMST), generating together 300 distinct EEG recording sessions. It was investigated whether NF training and monetary reward can enhance alpha suppression, consequently WM and whether any NF-training effect could be transferred to unrelated cognitive tasks. According to inter-individual variabilities in NF learning, it was also tested the effect of different mental strategies during maintenance. Hence, participants of Experiment I were instructed to perform a mental calculation task unrelated to the DMST, whereas participants of Experiment II to mentally rehearse the encoded visual object. Lastly, the effect of training was also explored on neighboring theta and beta frequency bands, related to memory processes. Results from Experiment I did show improved WM accuracy and reaction times (RTs) with a significant reward-anticipation effect. While no significant NF-training or reward-anticipation effects were found on enhancing alpha suppression, a reward-anticipation and NF effect was observed on theta suppression. Besides, a beta power decrease was unrelated to these factors. Results of Experiment II replicated improved WM performance with reward-anticipation effect, although neither NF-training nor reward-anticipation effects were found on oscillations and WM. Moreover, neither experiments demonstrated transfer effects of either WM or NF-training. Across both experiments, while during encoding enhanced WM accuracy over 5-days was found related to increased right parietal beta, faster RTs were found related to increased right parietal theta. During maintenance, increased right parietal beta was found to improve RTs, whereas increased left parietal beta was found to slow the performance. Despite a lack of NF-training benefits on WM, this exploratory analysis demonstrated how training-related improvements of WM are associated with oscillatory changes across training days. Based on these studies a proposal for an NF-training protocol was designed to explore declined cognitive-related oscillations in preclinical AD, providing the basis for a standard protocol to be further tested and developed.